Detection and localization of text in photorealistic images is a difficult, and not yet completely solved, problem. We propose the approach to solving this problem based on the method of semantic image segmentation. In this interpretation, text characters are treated as objects to be segmented. In this paper proposes the network architecture for text localization, describes the procedure for the formation of the training set, and considers the algorithm for pre-processing images, reducing the amount of processed data and simplifying the segmentation of the object “background”. The network architecture is a modification of well-known DeepLabv3 network and takes into account the specifics of images of text characters. The proposed method is able to determine the location of text characters in the images with acceptable accuracy. Experimental results of assessing the quality of text localization by the IoU criterion (Intersection over Union) showed that the obtained accuracy is sufficient for further text recognition.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Subtitle of host publicationConference proceedings
EditorsSanjay Misra, Osvaldo Gervasi, Beniamino Murgante, Elena Stankova, Vladimir Korkhov, Carmelo Torre, Eufemia Tarantino, Ana Maria A.C. Rocha, David Taniar, Bernady O. Apduhan
PublisherSpringer Nature
Pages825-834
Number of pages10
ISBN (Print)9783030243043
DOIs
StatePublished - 2019
Event19th International Conference on Computational Science and Its Applications, ICCSA 2019 - Saint Petersburg, Russian Federation
Duration: 1 Jul 20194 Jul 2019
Conference number: 19

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11622 LNCS

Conference

Conference19th International Conference on Computational Science and Its Applications, ICCSA 2019
Abbreviated titleICCSA 2019
Country/TerritoryRussian Federation
CitySaint Petersburg
Period1/07/194/07/19

    Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

    Research areas

  • Convolution neural network, Semantic segmentation, Text localization

ID: 47786980